Google Cloud introduces AI agent platform and new TPU chips at Next 2026

• Google Cloud introduced an AI agent platform and new TPUs at Next 2026.
• Google Cloud said 75% of customers use its AI, with new agent tools.

Google Cloud announced new AI products and infrastructure at Google Cloud Next 2026, including tools for building and managing AI agents. The company introduced updates across its platform, data systems, infrastructure, and security tools.

Google and Alphabet CEO Sundar Pichai said the company recorded 40% quarter-over-quarter growth in paid monthly active users of its Gemini Enterprise system in the first quarter of 2026. He said “every employee… can become a builder,” adding that organisations are shifting toward managing large numbers of AI agents.

CEO of Google Cloud, Thomas Kurian, said Gemini Enterprise connects enterprise data, applications, and AI agents. He described it as “the connective tissue between your data, your people, and all of your apps and AI agents,” used to integrate processes into a single system.

Google Cloud stated that nearly 75% of its customers are using its AI products. These include CIMB Niaga and FairPrice Group, as well as organisations including NASA, Mars Incorporated, and Merck.

CIMB Niaga has deployed AI agents built on the Gemini platform to support banking services in Indonesia. These agents provide financial guidance based on customer data.

FairPrice Group has integrated AI agents into its retail systems as part of its “Store of Tomorrow” programme. The agents are embedded in smart carts to connect in-store and digital interactions.

Mars is using Gemini Enterprise to automate workflows across marketing, research, and enterprise search. Merck is applying AI agents across research, clinical development, and commercial operations.

NASA is using AI agents to support flight readiness operations for the Artemis II mission. Virgin Voyages has deployed more than 1,000 AI agents, including systems used to reduce campaign production time by 40%.

Google Cloud reported that over the past year, 330 customers each processed more than one trillion tokens using its AI models, while 35 customers exceeded 10 trillion tokens. Its models now process more than 16 billion tokens per minute via API, up from 10 billion in the previous quarter.

Gemini Enterprise Agent Platform

The company introduced the Gemini Enterprise Agent Platform, which expands its Vertex AI system. Google Cloud said the platform acts as “a mission control for the Agentic Enterprise,” providing a central environment to build and manage AI agents.

The platform supports access to more than 200 AI models, including Gemini 3.1 Pro and models from partners such as Anthropic. It also adds support for Anthropic’s Claude Opus 4.7 model.

TPU infrastructure updates

Google Cloud announced its eighth-generation Tensor Processing Units, TPU 8t and TPU 8i, for training and inference workloads.

TPU 8t is built for model training and is intended to reduce development time through higher compute throughput and memory bandwidth. TPU 8i is designed for inference, with 288GB of high-bandwidth memory and 384MB of on-chip SRAM to support low-latency processing.

Google Cloud stated the new TPU systems deliver up to 80% better performance per dollar compared with the previous generation. Both chips are expected to become generally available later in 2026.

Data architecture and processing

The company introduced updates under its Agentic Data Cloud framework. Google Cloud said this enables “systems of action” that allow enterprise data to be used in real time by AI agents.

A cross-cloud lakehouse based on Apache Iceberg allows organisations to query data stored in platforms such as AWS and Azure without transferring it. The system provides zero-copy access to enterprise applications and data platforms.

Google Cloud also introduced a serverless processing engine for Apache Spark, designed to run up to 4.5 times faster than open-source alternatives. A Data Agent Kit enables development of workflows using natural language inputs across Python, Spark, and SQL.

A Knowledge Catalog feature creates a unified data context layer across enterprise systems, allowing AI agents to access structured and unstructured data.

Security and threat detection

Google Cloud outlined updates to its cybersecurity platform, combining its Threat Intelligence and Security Operations tools with the Wiz cloud security platform.

The company reported that its Triage and Investigation Agent has processed more than five million alerts, reducing analysis time from 30 minutes to about 60 seconds. Wiz now supports platforms including Databricks and multiple AI agent development environments.

Google Cloud introduced Dark Web Intelligence, which analyses external threat data using AI models. It also introduced a Threat Hunting Agent designed to identify new attack patterns.

Additional tools automate detection rule creation and support incident investigation and remediation.

Google Cloud also introduced Google Cloud Fraud Defensereplacing reCAPTCHA. The company said it is designed to “discern the legitimacy and authorisation of bots, humans, and AI agents” during online interactions.

Workspace and productivity tools

The company announced updates to Google Workspace under a system called Workspace Intelligence. This layer connects data across emails, documents, and meetings to support AI-driven task execution.

New features include AI-assisted email management in Gmail, automated task execution across Workspace applications, and tools that generate documents and presentations based on user data. A Google Drive Projects feature organises files and communications into project-based workflows.

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information.

AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.